Integrated analysis of bulk RNA and single-cell RNA sequencing data reveals potential biomarkers and immune infiltrates associated with N7-methylguanosine in osteoarthritis

Abstract

Objective:

The N7-methylguanosine (m7G) modification is known as a common post-transcriptional modification of RNA that has been found to be involved in the pathogenesis of various diseases. However, its role in osteoarthritis (OA) remains largely unknown. This study aimed to identify the genes associated with m7G modification in OA and further investigate their diagnostic value and immune infiltrates.

Methods:

We constructed an OA risk model based on key m6A regulators using LASSO regression. Unsupervised clustering analysis was used to detect diverse m7G modification phenotypes and m7G gene clusters based on key m7G regulators in the 106 OA samples. We evaluated the features of the immune microenvironment between distinct risk groups, m7G modification phenotypes, and m7G gene clusters, using immune infiltration analysis. Single-cell analysis was performed to determine the relationship between the key m7G regulators and cartilage degeneration. Finally, in vitro and in vivo experiments were conducted to explore the potential roles of key m7G regulators in OA.

Results:

Four key m7G regulators, including EIF1, JUND, NUDT16L1, and NCBP1, and two risk groups, m7G modification phenotypes, and m7G gene clusters were identified. A nomogram model constructed based on the key m7G regulators demonstrated its predictive role in the occurrence of OA. Moreover, the risk score of the patients in the high-risk group was higher than that of the patients in the low-risk group. Different immune infiltration characteristics were found among the risk groups, m7G modification phenotypes, and m7G gene clusters. Additionally, single-cell analysis confirmed ubiquitous expression of EIF1 and JUND across all chondrocyte subtypes. Furthermore, both genes were downregulated in OA cartilage, and ROC analysis demonstrated that EIF1 and JUND exhibited excellent performance in OA diagnosis. Finally, it was confirmed that inhibiting the expression of EIF1 and JUND induced upregulation of MMP13 and downregulation of collagen II. Further in vivo experiments indicated that suppression of EIF1 and JUND contributed to the progression of OA.

Conclusion:

Our study revealed four m7G regulators, including EIF1, JUND, NUDT16L1, and NCBP1, as novel biomarkers that may be associated with the immune infiltration during the progression of OA. Moreover, EIF1 and JUND may serve as potential therapeutic targets for preventing cartilage degeneration. However, further experiments are required to elucidate the molecular mechanisms underlying OA.

Introduction

Osteoarthritis (OA), a common degenerative disease, mostly affects middle-aged and elderly populations and is characterized by joint pain, deformity, and disability^[1]. The occurrence of OA has been associated with multiple factors, including but not limited to aging, sex, obesity, trauma, abnormal limb alignment, and genetic predisposition^[2]. The cartilage, bone, synovium, ligament, muscle, and periarticular fat undergo remarkable pathological changes, leading to cartilage degeneration, synovial inflammation, angiogenesis, subchondral osteosclerosis, and osteophyte formation^[3]. Although the mechanisms underlying OA have not yet been fully elucidated, numerous regulatory phenotypes, including cell death, oxidative stress, cell senescence, epigenetic modification, inflammatory response, and immunoregulation, have been implicated in the initiation and progression of OA^[4–8]. Owing to the complexity and diversity of OA pathogenesis, there is currently a lack of effective therapeutic strategies to prevent OA progression. Therefore, timely diagnosis using reliable biomarkers is crucial for the holistic management of OA. Currently, the majority of diagnostic biomarkers are developed based on the analysis of cartilage, synovial tissue, and subchondral bone^[9–11]. However, collecting these specimens for testing often requires invasive procedures, which can cause pain and increase the risk of infection. Furthermore, owing to the diverse pathological changes in OA progression, there may be deviations and limitations in predicting the occurrence of OA using biomarkers of one pathological change. Peripheral blood is widely recognized for its non-invasive nature, ease of acquisition, and suitability for disease tracking and follow-up. Consequently, numerous blood-derived biomarkers have been developed to predict the occurrence and prognosis of various diseases including cancer, Alzheimer’s disease, and stroke^[12–14]. Nevertheless, effective blood-derived biomarkers for OA diagnosis are lacking.

Recently, the understanding of OA has advanced from a purely mechanical disorder resulting from cartilage injury to a complex biological response involving inflammation and immune processes^[8,15]. The T and B cell profiles are altered in patients with OA, indicating that the acquired immune system may play a critical role in the progression of OA^[16]. Moreover, as a central regulator of innate immunity, M2 macrophages can maintain a favorable anti-inflammatory and prechondrogenic environment for chondrocytes in an inflammatory OA microenvironment^[17,18]. Available evidence suggests a strong association between immune cells, immune-related signaling pathways, and mechanisms of OA occurrence. Therefore, screening biomarkers from an immunological perspective may help to diagnose OA in a timely manner and provide therapeutic targets.

RNA methylation, including that of N1-methyladenosine (m1A), N6-methyladenosine (m6A), 5-methylcytosine (m5C), and 7-methylguanosine (m7G), is a hallmark of epigenetics^[19]. The m7G modification is a prevalent post-transcriptional modification among these RNA methylations, commonly observed in tRNA, rRNA, miRNA, and the 5′ cap of mRNA^[20–22]. Moreover, the m7G modification participates in multiple aspects of RNA metabolism, including RNA processing, stabilization, maturation, and translation^[23]. An increasing number of studies have demonstrated that m7G modification plays an important role in regulating the pathogenesis of various diseases such as cancers, brain disorders, and autoimmune diseases^[23–25]. Relevant investigations have been conducted to explore the correlation between m7G modification and OA. For instance, Chen et al comprehensively explored the dysregulation of m7G modification in the immune landscape during the progression of OA and four m7G-related hub genes, SNUPN, METTL1, EIF4E2, and CYFIP1, as novel biomarkers for predicting the occurrence of OA^[26]. Hao et al also identified four m7G biomarkers, including DCP2, EIF4E2, LARP1, and SNUPN, in the development of OA, and an m7G-related scoring model associated with the immune microenvironment was constructed based on these four biomarkers^[27]. Remarkably, the current findings regarding the role of m7G modification in OA progression have been primarily derived from transcriptomic changes observed in synovial tissues^[11,26,27]. However, there is still unknown about the m7G modification-related signature in the peripheral blood of OA patients.

HIGHLIGHTS

• Our study has revealed four key m7G regulators, including EIF1, JUND, NUDT16L1, and NCBP1, as novel biomarkers in the progression of osteoarthritis (OA).

• All key m7G regulators may be involved in the immune infiltration of OA.

• In addition, two key M7G regulators, including EIF1and JUND, may serve as potential therapeutic targets for preventing cartilage degeneration.

Our study has been reported in line with the TITAN criteria^[28]. In the current study, a variety of bioinformatics methods were employed to investigate the m7G modification-related signature using bulk RNA sequencing (RNA-seq) data from the peripheral blood of normal individuals and OA patients, and to further determine the relationship between m7G modification and immune infiltrates in OA. We identified four key m7G regulators as biomarkers and constructed a predictive model and a risk model based on these genes. Furthermore, external single-cell RNA-seq (scRNA-seq) and RNA-seq datasets of cartilage confirmed the expression of key m7G regulators in articular cartilage and uncovered their potential roles in OA pathogenesis. These findings may provide novel biomarkers and potential targets for early diagnosis and treatment of OA.

Materials and methods

The study design is illustrated in Figure 1. Briefly, GSE48556 and m7G regulator gene sets were used to identify the differentially expressed m7G regulators. Multiple bioinformatics methods, including weighted gene co-expression network analysis (WGCNA), LASSO, Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and single-sample gene set enrichment analysis (ssGSEA), were used to screen key m7G regulators and further investigate their biological functions, pathways, diagnostic values, and immune microenvironment features of key m7G regulators. Additionally, both scRNA-seq and RNA-seq cartilage data were used to determine the relationship between key m7G regulators and cartilage degeneration. Finally, quantitative real-time PCR (qRT-PCR) and western blot (Wb) were conducted to explore the potential roles of key m7G regulators in OA.

Figure 1.

The flow chart of the study.

Data collection and processing

The GSE48556 dataset obtained from the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) database contains 106 blood samples from patients with OA and 33 blood samples from healthy individuals without OA, which were used to screen key m7G regulators and investigate their associated biological functions, pathways, and immune infiltrates. The GSE104782 dataset includes the scRNA-seq data of 10 osteoarthritic cartilage samples, which contain 1600 cells, and was used to validate the expression of key m7G regulators in cartilage. Additionally, the GSE114007 dataset containing 18 normal and 20 OA knee cartilages was used as an external validation cohort to determine the difference in key m7G regulators between normal and OA cartilages.

We downloaded the m7G modification-related regulators from the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb). Differentially expressed genes (DEGs) between OA and normal samples were screened using the R package “limma” (version 3.56.2) with P-value < 0.05. Subsequently, common genes between DEGs from the GSE48556 dataset and previously obtained m7G regulators were considered as differentially expressed m7G regulators in OA. The “ggplot2” package (version 3.4.3) was adopted to visualize the differentially expressed m7G regulators. Moreover, the relationship among m7G regulators in patients with OA was analyzed using the R package “corrplot” (version 0.92) and Spearman’s correlation analysis, and the cutoff criterion was P-value < 0.05. The m7G regulators with significant correlations were visualized using the “ggMarginal” (version 0.10.1) and “ggplot” (version 3.4.3) packages of the R software.

WGCNA and LASSO analysis

The R package “WGCNA” (version 1.72.1) was employed to analyze all DEGs in the GSE48556 dataset and to determine the soft thresholding power, aiming to narrow down the range of hub genes. Subsequently, a gene co-expression network was built, and the genes were grouped into several modules with distinct colored labels. The OA trait was evaluated using Pearson’s correlation coefficient, and the top three modules with the highest correlation with the OA trait were identified as essential modules. The genes located in the essential modules were intersected with the differentially expressed m7G regulators to obtain hub m7G regulators that were significantly associated with OA clinical phenotypes. The threshold points for hub genes in each module were |GS| > 0.2 and |MM| > 0.8. LASSO analysis with turn/penalty parameters was performed to further screen for key m7G regulators. The best regularization parameters were selected using the R package “glmnet” (version 4.1.8), and 10-fold cross-validation was performed to fit the LASSO model. Finally, common genes after the intersection of WGCNA, LASSO, and differentially expressed m7G regulators were identified as key m7G regulators.

Functional enrichment analysis of key m7G regulators

To investigate potential biological functions and associated pathways of key m7G regulators, we conducted GO and KEGG pathway enrichment analyses using the EnrichR database (https://maayanlab.cloud/Enrichr/). P-value < 0.05 was determined as a significant threshold of analyses.

Construction of the nomogram model and risk model in OA

The nomogram model was developed based on key m7G regulators using the “rms” (version 6.7.1) package of R software to predict the occurrence of OA. A calibration plot was used to assess the concordance between projected and actual values. The decision and clinical impact curves were plotted to assess the potential of the model to provide valuable insights into illness prediction in patients with OA. Additionally, a risk model was constructed to investigate the correlation between key m7G regulators and OA progression using the LASSO regression. The Risk Score was calculated using the following formula: Risk Score = ${\sum }^i$ Coefficients_i × Expression level of signature_i.

Immune cell infiltration analysis

Patients with OA in the GSE48556 dataset were classified into high-risk and low-risk groups based on their risk scores. To perform ssGSEA, we collected genes associated with 28 distinct immune cell types from relevant literature^[29]. Next, the R package “GSVA” (version 1.48.3) was used to quantify immune cells in both high-risk and low-risk groups. The determining criterion for a significant difference was P-value of < 0.05. Additionally, the R package “psych” (version 2.4.3) was used to perform the correlation analysis and significance testing to elucidate the relationship between immune cells and key m7G regulators (P-value < 0.05).

Identification and construction of m7G modification phenotypes and scores in OA

Unsupervised clustering analysis was used to identify distinct m7G modification phenotypes in OA patients. The R package “ConsensusClusterPlus” (version 1.64.0) was used to divide the patients with OA into different clusters. Principal component analysis (PCA) was performed to evaluate m7G modification phenotypes. Principal components (PC) 1 and 2 were extracted to calculate the signature scores. Subsequently, signature scores were used to establish the m7G score as follows: m7G score = ${\sum }^{\left(PC1i+PC2i\right)}$. ssGSEA was used to quantify immune cell levels between the two m7G modification phenotypes. Additionally, differential analysis was performed to identify DEGs between samples from distinct m7G modification phenotypes using the R package “limma” (version 3.56.2) with the criteria set as log₂ |fold change (FC) | > 0.5 and P < 0.05. DEGs were further used for constructing m7G gene clusters through “ConsensusClusterPlus” (version 1.64.0). Consistent with the aforementioned methods, the signature score and immune cell infiltration analysis were calculated based on m7G gene clusters.

Validation of the expression of key m7G regulators in cartilage

The GSE140782 dataset was filtered using the R package “Seurat4.0” (version 4.0) with the restriction of nCountRNA 25000, followed by quality control analysis of the gene expression matrix to identify single-cell hypervariable genes. The samples were subjected to PCA for cluster selection, followed by cell clustering using the “FindClusters” function in the R package “Seurat” (version 4.0). Finally, annotation was performed using SingleR (https://comphealth.ucsf.edu/app/singler). Additionally, the R package “FeaturePlot” (version 4.3.0) was used to annotate the key m7G regulators.

The GSE114007 dataset was used as an external validation dataset to further corroborate the expression levels of key m7G regulators in OA and normal cartilages. Moreover, the receiver operator characteristic (ROC) curve was plotted using the R package “timeROC” (version 0.4), and the area under the curve (AUC) was calculated to assess the diagnostic efficacy of key m7G regulators in OA. Genes with AUC > 0.7 and P < 0.05, contained favorable predictive ability.

Cartilage sample collection

This study was designed in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Affiliated Hospital of Guizhou Medical University (No. 276 in 2022). Osteoarthritic cartilage was obtained from the knee joints of six patients undergoing total knee arthroplasty, while normal cartilage was acquired from the hip joints of six patients with no previous history of OA or rheumatoid arthritis who underwent total hip arthroplasty or artificial femoral head arthroplasty for hip fractures. X-Ray images used in this study have been provided with patient consent for publication. Clinical information of the patients is presented in Supplemental Digital Content Table S1, available at http://links.lww.com/JS9/F256.

Safranin O/fast green staining

A fraction of each cartilage sample was fixed with 4% paraformaldehyde for 48 h, and decalcified with 10% EDTA at 37°C for 2 weeks. After embedding in paraffin, cartilage samples were sectioned at 5 μm thickness in the sagittal plane. Subsequently, the sections were stained with safranin O/fast green (Servicebio, C1371) according to the manufacturer’s instructions. The Osteoarthritis Research Society International (OARSI) score system was applied to evaluate the severity of cartilage degeneration in destabilization of the medial meniscus (DMM) OA mouse model.

Quantitative real-time PCR

Total RNA was isolated and qRT-PCR was performed as described previously^[9]. In brief, total RNA of cartilage was isolated using Triquick Reagent (Solarbio, China, R1100) according to the manufacturer’s instructions. PrimeScript™ RT Master Mix (Takara Bio Inc., Japan, RR036A) was used to obtain the cDNA. TB Green® Premix Ex Taq™ II (Takara Bio Inc., Japan, RR820A) was used for PCR detection. Primer sequences are listed in Supplemental Digital Content Table S2, available at http://links.lww.com/JS9/F257.

Cell culture and treatment

Human C28/I2 chondrocytes were obtained from the Shanxi Key Laboratory of Bone and Soft Tissue Injury Repair, Second Hospital, Shanxi Medical University, China. DMEM/F12 medium (Gibco, C11330500BT) containing 10% fetal bovine serum (Gibco, 10099141), 1% penicillin, and 1% streptomycin sulfate was used for the cell culture. The cells were then placed in a 37°C incubator with 5% CO₂. The culture medium was refreshed every other day. When the cell density reached 60%–70%, the C28/I2 chondrocytes were exposed to 50 µM t-butylhydroperoxide (TBHP) (Sigma-Aldrich, USA, 458139) for 4 h to establish an OA cell model, as previously reported^[30,31]. Subsequently, the total protein was extracted for the Wb analysis.

Cell transfection with small interfering RNAs

A total of 2 × 10⁵ C28/I2 chondrocytes were cultured in 6 cm dishes. After seeding and adhering for 24 h, chondrocytes reached 70%–80% confluency and were then transfected with small interfering RNAs (siRNAs) or a negative control (si-NC) (Sangon Biotech, China) using Lipofectamine 8000 (Beyotime, China, C0533) according to the manufacturer’s instructions. Total protein was collected after chondrocytes were treated with siRNAs or si-NC for 72 h.

Wb assay

The total protein of C28/I2 chondrocytes was extracted using RIPA lysis buffer (Proteintech, China, PR20035) containing 1% phenylmethylsulfonyl fluoride (PMSF, Solarbio, 329-98-6) for 30 min on ice. Subsequently, cell lysates were harvested and collected by centrifugation at 12 000 rpm for 15 min at 4°C. The supernatant was then acquired and the protein concentration was quantified using a BCA assay kit (Solarbio, China, PC0020). After diluted with 5× SDS-PAGE loading buffer (Epizyme Biotech, China), the proteins were boiled at 95°C for 10 min. A total of 10 µg of total protein from each sample was separated using 7.5% or 10% SDS-PAGE gel (Epizyme Biotech, China) and subsequently electroblotted onto a 0.2 or 0.45 μm PVDF membranes (Millipore, USA). After blocking with 5% skim milk at room temperature for 2 h, the membranes were incubated with anti-EIF1 (Proteintech, China, 15276-1-AP; 1:1000), anti-JUND (HUABIO, China, ET1612-92;1:2000), anti-MMP13 (Servicebio, China, GB11247-1-100; 1:1000), anti-collagen II (Servicebio, China, GB11021-100; 1:2000), anti-collagen II (Abcam, USA, ab188570; 1:1000), anti-α-tubulin (Proteintech, China, 11224-1-AP; 1:50 000), and anti-GAPDH (Proteintech, China, 10494-1-AP; 1:10 000) overnight at 4°C. The next day, membranes were incubated with secondary antibodies for 2 h at room temperature. Protein bands were acquired using a ChemiScope 6000 (CLINX, China), and the density of each band was quantified using ImageJ software.

Animal experiments

The protocol of animal experiments in this study was approved by the institutional Animal Care and Use Committee of Guizhou Medical University (No.2201177). We performed the surgery of DMM in the left knee joint to construct the experimental OA model, and the corresponding right knee joint was undergoing sham surgery. Eight-week-old-male C57BL/6J mice weighing 20–25 g were purchased from the Experimental Animal Center of Guizhou Medical University , and randomly divided into three groups (n = 5): sham+ si-NC, DMM+ si-NC, DMM+ si-EIF1(or si-JUND). Both siRNA and the corresponding si-NC were injected into their knee (5 nmol; 10 µL) once a week after surgery. Mice were sacrificed at 8 weeks after DMM or sham surgery. Knee samples were collected and fixed in 4% paraformaldehyde for further investigation. Safranin O/fast green staining and OARSI score system were applied to evaluate the severity of cartilage degeneration in deferent groups. In addition, this work of animal experiments has been reported in accordance with the ARRIVE guidelines (Animals in Research: Reporting In Vivo Experiments)^[32].

Statistical analysis

R software (version 4.3.1), along with the relevant packages, was used to perform the bioinformatics analyses. Comparisons of the bioinformatic analyses were performed using the Wilcoxon test. Normally distributed measurement data from triplicate independent experiments are presented as mean ± standard error of the mean. GraphPad Prism software (version 8.4.0) was used for statistical analyses. Statistical significance was assessed using unpaired Student’s t-test. Differences among groups were compared using one-way analysis of variance (ANOVA) followed by Dunnett’s t-test. P significance was set at P < 0.05.

Results

Expression of m7G regulators in OA and normal samples

In total, 59 m7G regulators were obtained from MSigDB. After intersection with the DEGs from the GSE48556 dataset, 16 m7G regulators with significant differences were identified (Fig. 2A and B). Nine m7G regulators (NUDT16L1, CYFIP2, PARN, NCBP1, EIF4G1, SNUPN, NUDT5 and EIF4E2) were upregulated in OA patients compared with healthy individuals. Conversely, seven m7G regulators (EIFI, LSM1, LENG8, TNPO1, RPS12, EIF4G3, and JUND) were downregulated in OA patients. The location of these differentially expressed m7A regulators on the chromosomes are shown in Figure 2C and Supplemental Digital Content Table S3, available at http://links.lww.com/JS9/F258.

Figure 2.

Expression of m7G modification-related regulators in OA patients and normal individuals. (A) Histogram of the differences in expression of m7G regulators between normal individuals and OA patients. (B) Heatmap of expression levels of m7G regulators between normal individuals and OA patients. (C) The location of m7G regulators on human chromosomes. (D) The correlations among m7G regulators in OA.

Correlations among the significantly altered m7G regulators were further investigated in OA samples. Many m7G regulators were strongly correlated with each other (Fig. 2D). For instance, SNUPN expression was positively associated with NUDT16L1 and EIF4G1, whereas the expression levels of EIF4G3 and TNPO1 were negatively correlated with EIF4E2. Additionally, high expression level of NUDT5 was positively correlated with EIF4E2 (Supplemental Digital Content Figure S1, available at http://links.lww.com/JS9/F255).

Identification of key m7G regulators in OA

All DEGs in the GSE48556 dataset were used to construct a gene co-expression network. Sample cluster analysis was performed to group OA samples and eliminate outliers (Fig. 3A). Next, a threshold of 4 and a power of 13 were set to construct an adjacency matrix using the correlation coefficient between any two genes (Fig. 3B). Average linkage hierarchical clustering was performed to identify modules consisting of closely connected genes, resulting in eight color-coded modules. Unassigned genes are represented in grey (Fig. 3C). The correlation between each module and OA was calculated and visualized, leading to the identification of a key module composed of three merged modules containing a total of 451 essential genes colored purple, black, and green-yellow (Fig. 3D). After intersection these essential module genes with the above-mentioned 16 m7G regulators, 5 hub m7G regulators including EIF1, JUND, NCBP1, SNUPN, and NCDT16L1 were then identified. In addition, LASSO analysis was employed to perform ten-fold cross-validation and further screen for key m7G regulators (Fig. 3E and F). Ultimately, the integration of WGCNA and LASSO analyses, in conjunction with the previously identified 16 m7G regulators, led to the identification of a set of four pivotal genes for m7G modification in OA, namely EIF1, JUND, NCBP1, and NUDT16L1 (Fig. 3G).

Figure 3.

Identification of key m7G regulators in OA using WGCNA and LASSO analysis. (A) Sampling clustering of the training set. (B) Analysis of soft threshold power to match the scale-free topology model and the average connectivity of soft threshold power; 4 was chosen as the value for building the scale-free network. (C) Tree of gene modules. (D) Heatmap of gene module and trait association. (E) coefficient index for the LASSO path. (F) Parameters of LASSO logistic algorithm. (G) Venn diagram shows the shared genes among the top three significant module genes in WGCNA, differentially expressed m7G regulators and LASSO analysis.

Biological functions and pathways of key m7G regulators

The EnrichR database was used to perform GO functional and KEGG pathway enrichment analyses of the key m7G regulators. GO enrichment analysis showed that key m7G regulators were mainly involved in RNA catabolic processes, regulation of translational initiation, and mRNA metabolic processes in terms of biological process (BP); nucleus, intracellular membrane-bounded organelle in terms of cell component (CC); and RNA binding, RNA m7G cap binding, and translation initiation factor activity in terms of molecular function (MF) (Fig. 4A). The GO enrichment results were visualized using an enrichment circle chart (Fig. 4B). These key m7G regulators were mostly enriched in GO terms 006417 (BP), 0043231 (CC), and 0003723 (MF), which correlated with the regulation of translation, intracellular membrane-bound organelles, and RNA binding. Figure 4C illustrates the enriched BPs associated with each key m7G regulator. Through KEGG analysis, we found that key m7G regulators were mostly enriched in the RNA transport, mRNA surveillance, and IL-17 signaling pathways (Fig. 4D).

Figure 4.

Biological functions and pathways of key m7G regulators. (A) GO analysis of key m7G regulators from the viewpoints of the biological process (BP), cellular component (CC), and molecular function (MF). (B) The circle diagram of gene enrichment numbers for each GO item. (C) The BPs associated with each key m7G regulator. (D) KEGG analysis of key m7G regulator.

Construction of the nomogram model and risk model in OA

To explore the predictive value of key m7G regulators for OA occurrence, a nomogram model was established using logistic analysis. As shown in Figure 5A, the expression levels of these four key m7G regulators were negatively correlated with OA probability. The ROC and calibration curves exhibited high predictive capability for OA (Fig. 5B and C). According to the Decision Curve Analysis (DCA) curve, the red line consistently remained above both the gray and black lines within the range of 0–0.8, indicating that this nomogram model has the potential to provide valuable assistance in predicting patients with OA (Fig. 5D). Furthermore, the clinical impact curve illustrated the superior predictive ability of the line-plot model (Fig. 5E).

Figure 5.

Construction of a nomogram predictive model in OA. (A) Establishment of the nomogram using 4 key m7G regulators. (B) The ROC curve analysis shows the predictive efficiency of the nomogram. (C) The calibration curve used to assess the predictive accuracy of the nomogram (the closer to the ideal dotted line, the more reliable the result). (D) The decision curve of the nomogram (the farther the endpoint of the red line is from the gray line, the higher the accuracy is). (E) The clinical impact curve used to determine the clinical utility of the nomogram (the red line represents the model’s prediction of high-risk patients, while the blue dotted line indicates the actual high-risk patients).

To further investigate the biological significance of these four key m7G regulators, LASSO analysis was used to construct a risk model for OA: Risk score = (−2.993975936 × EIF1) + (−0.039036604 × JUND) + (2.510394342 × NCBP1) + (3.282484463 × NUDT16L1). Risk scores were initially calculated for normal and OA samples, revealing a significantly higher risk score in patients compared than in healthy individuals (Fig. 6A). Subsequently, all OA samples were randomly divided into two distinct subsets based on risk scores at a ratio of 1:1, thereby establishing high- and low-risk groups. The results indicated that patients in the high-risk group had a significantly higher risk score than those in the low-risk group (Fig. 6B). Moreover, the AUC value was 0.765, indicating exceptional efficacy of this risk model (Fig. 6C). Interestingly, the expression of key m7G regulators was significantly different between risk groups (Fig. 6D and E).

Figure 6.

Construction of a risk model in OA. (A) Comparison of risk scores between normal and OA patients. (B) Comparison of risk scores in OA patients between the high-risk group and the low-risk group. (C) The ROC curve analysis shows the predictive efficiency of the risk model. (D–E) Heatmap and histogram depict the expression of key m7G regulators expression between risk groups in OA. **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Immune infiltration features of m7G modification-related risk groups

We further explored the differences in immune infiltration between the risk groups using ssGSEA analysis. The results showed a significant infiltration of nine immune cells in the high-risk group, including activated B cells, activated CD8 T cells, activated dendritic cells, effector memory CD8 T cells, immature B cells, macrophages, regulatory T cells, type 1 T helper cells, and type 2 T helper cells. The levels of seven immune cells, including activated CD4 T cells, central memory CD4 T cells, mast cells, MDSC, natural killer cells, natural killer T cells, neutrophils, and type 17 T helper cells, were found to be increased in the low-risk group (Fig. 7A). Moreover, key m7G regulators also observed to exhibit correlation with immune cell infiltration (Fig. 7B). For instance, a positive correlation was identified between EIF1 and MDSC, JUND, central memory CD4 T cells, NCBP1 and CD8 T cells, and NUDT16L1 and type 1 T helper cells. Nevertheless, a negative correlation was found between EIF1 and type 2 helper cells, JUND and macrophages, NCBP1 and mast cells, NUDT16L1 and activated CD4+ T cells.

Figure 7.

Immune infiltration features in risk groups of OA. (A) Differences of immune cell infiltration between risk groups. *P < 0.05, **P < 0.01, and ***P < 0.001. (B) Correlations between key m7G regulators and immune cell infiltration. *P < 0.05, **P < 0.01, and ***P < 0.001.

Identification of m7G modification phenotypes and m7G gene clusters in OA

According to the expression levels of the four key m7G regulators, unsupervised cluster analysis was conducted to detect m7G modification phenotypes in OA patients from the GSE48556 dataset, and two distinct clusters were identified with an optimal k value of 2 (Fig. 8A). PCA showed that patients with OA could be divided into two subtypes (Fig. 8B). There were 47 patients with OA in Cluster A and 59 in Cluster B. The immune microenvironment features of the two distinct m7G modification phenotypes were also explored using ssGSEA analysis. A higher infiltration level of activated CD4 T cells, central memory CD4 T cells, MDSC, natural killer cells, natural killer T cells, neutrophils, T follicular helper cells, and type 17 T helper cells was found in m7G cluster A, while a higher infiltration level of activated CD8 T cells, activated dendritic cells, effector memory CD4 T cells, immature B cells, macrophages, regulatory T cells, type 1 T helper cells, and type 2 T helper cells was mainly associated with m7G cluster B (Fig. 8D).

Figure 8.

Identification of m7G modification phenotypes and m7G gene clusters in OA. (A) Two m7G modification phenotypes were identified using unsupervised consensus cluster analysis (k = 2). (B) PCA analysis of two m7G clusters. (C) Comparison of m7G scores between m7G cluster A and B. (D) Differences of immune cell infiltration between m7G clusters. *P < 0.05, **P < 0.01, and ***P < 0.001. (E) Differences of immune cell infiltration between gene clusters. *P < 0.05, **P < 0.01, and ***P < 0.001. (F) Two gene clusters were identified using unsupervised consensus cluster analysis (k = 2). (G) PCA analysis of two m7G gene clusters. (H) Comparison of m7G scores between gene cluster A and B.

Subsequently, 193 DEGs were identified between the two m7G modification phenotypes, which were considered m7G phenotype-related genes. We performed unsupervised consensus cluster analysis based on the expression of m7G modification phenotype-related genes in patients with OA. Interestingly, patients were also categorized into two distinct gene clusters (Fig. 8F and G). Immune cell infiltration differed significantly between the two m7G gene clusters (Fig. 8E). Cluster A showed higher infiltration levels of activated CD4 T cells, CD56bright natural killer cells, central memory CD4 T cells, MDSC, monocytes, natural killer cells, natural killer T cells, neutrophils, T follicular helper cells, and type 17 T helper cells. Cluster B showed higher infiltration levels of activated CD8+ T cells, activated dendritic cells, effector memory CD4+ T cells, immature B cells, immature dendritic cells, macrophages, memory B cells, plasmacytoid dendritic cells, regulatory T cells, type 1 T helper cells, and type 2 T helper cells. Finally, the m7G modification scores of the m7G modification phenotypes and m7G gene clusters were calculated using PCA. Significantly, a higher m7G modification score was found in m7G cluster B and gene cluster A (Fig. 8C and H).

Validation of the expression of key m7G regulators in cartilages

To determine the potential roles of key m7G regulators in cartilage degeneration, we used the scRNA-seq dataset GSE104782 to validate their expression of key m7G regulators in cartilages. Highly variable genes were initially identified by single-cell analysis of chondrocytes obtained from OA cartilages. The top 10 variations were selected and labeled accordingly (Supplemental Digital Content Figure S2A, available at http://links.lww.com/JS9/F255). The scRNA-seq data were normalized and 20 principal components were selected for subsequent analysis (Supplemental Digital Content Figure S2B, available at http://links.lww.com/JS9/F255). The “Seurat” package was employed for dimensionality reduction and batch effect removal.

Subsequently, unsupervised analysis employing the uniform manifold approximation and projection (UMAP) method was conducted to cluster the cells together. As a result, a total of 13 cell clusters were identified with distinct molecular profiles (Supplemental Digital Content Figure S2C, available at http://links.lww.com/JS9/F255). A heat map was constructed for each cluster to visualize gene expression phenotypes (Fig. 9A). Finally, cellular annotation was performed using the “singleR” package to accurately assign the cell subtypes. Results showed that all cell subtypes were classified as chondrocytes (Supplemental Digital Content Figure S2D, available at http://links.lww.com/JS9/F255). As shown in Figure 9B–E, the expression of EIF1 and JUND was widespread across all subtypes of chondrocytes, whereas NUDT16L1 and NCBP1 exhibited limited expression.

Figure 9.

Validation of the expression of key m7G regulators in cartilages. (A) Heatmap shows the clusters in GSE140782 using scRNA-seq analysis. (B–E) Histogram depicts the expression of key m7G regulators in subtypes of chondrocytes, including EIF1 (B), NUDT16L1 (C), JUND (D), and NCBP1 (E). (F) The expression of key m7G regulators between normal and OA cartilages in GSE114007 dataset. *P < 0.05, and **P < 0.01. (G) The ROC curve analysis shows the diagnostic values of key m7G regulators in the GSE114007 dataset.

In addition, the external RNA-seq dataset GSE1140007 was used to validate the expression of key m7G regulators in normal and OA cartilages. The levels of EIF1 and JUND were significantly lower in OA cartilage than those in normal cartilage, whereas no significant differences were observed in the expression of NUDT16L1 and NCBP1 (Fig. 9F). Moreover, the AUC values for EIF1, JUND, NUDT16L1, and NCBP1 were 0.706, 0.783, 0.589, and 0.536, respectively (Fig. 9G). These findings indicate that EIF1 and JUND may have promising diagnostic performance for OA.

EIF1 and JUND were identified as essential regulators in cartilage degeneration

Furthermore, we obtained cartilage samples to validate the findings of the bioinformatics analyses (Fig. 10A). Safranin O/fast green staining confirmed the degradation of OA cartilage (Fig. 10B). Subsequently, we conducted qRT-PCR to detect differences in EIF1 and JUND between OA and normal cartilages. qRT-PCR results showed that there was a significant decrease in the mRNA expression levels of EIF1 and JUND in OA cartilage compared to those in normal cartilage (Fig. 10C and D).

Figure 10.

Validation the expression of EIFI and JUND between normal and OA cartilages using qRT-PCR. (A) X-rays and intraoperative images of knee and hip joint from an OA patient (a 57-year-old female) and a relatively health donor (a 68-year-old female). (B) Safranin O/fast green staining of normal and OA cartilages. Scale bar, 1.0 mm. (C–D) The mRNA expression of EIFI (D) and JUND (E) between normal (n = 6) and OA (n = 6) cartilages using qRT-PCR analysis. **P < 0.01.

TBHP was used to establish the OA cell model. As shown in Figure 11A and B, the expression of MMP13 was upregulated, whereas that of collagen II was downregulated in TBHP-induced chondrocytes. Moreover, the protein level of EIF1 was significantly decreased, whereas JUND was elevated in TBHP-treated chondrocytes compared to that in normal controls. To further investigate the potential roles of EIF1 and JUND in cartilage degeneration, we knocked them down using distinct sequence-specific siRNAs. First, a WB assay was performed to confirm the knockdown efficiency of EIF1. As shown in Figure 11C and D, the expression of EIF1 was remarkably decreased when chondrocytes were transfected with si-EIF1-1 and si-EIF1-2. Moreover, the knockdown efficiency of si-EIF1-2 was higher than that of si-EIF1-1 alone. Hence, we used si-EIF1-2 in subsequent experiments. After transfection of chondrocytes with si-EIF1-2, the expression of MMP13 increased, while the level of collagen II decreased (Fig. 11F). We also silenced JUND using three different siRNAs, and the results revealed that all siRNAs reduced the expression of JUND in the chondrocytes (Fig. 11C and E). Furthermore, si-JUND-2 showed the best knockdown efficiency; therefore, si-JUND-2 was used in subsequent functional experiments. As shown in Figure 11G, the expression of collagen II was downregulated, whereas that of MMP13 was upregulated when chondrocytes were transfected with si-EIF1-2 and si-JUND-2. In addition, we administered specific EIF1 siRNA and JUND siRNA respectively to an DMM-induced OA mouse model to verify the potential roles of EIF1 and JUND in the development of OA. Safranin O/fast green staining was conducted 8 weeks after surgery (Supplemental Digital Content Figure S3 A and C, available at http://links.lww.com/JS9/F255). The results showed that the OARSI core of the si-EIF1 group was significantly higher than those of Sham and DMM groups. Meanwhile, similar results were found in the si-JUND group. These findings suggested the potential protective roles of EIF1 and JUND in the progression of OA.

Figure 11.

In vitro experiments exploring the potential role of EIFI and JUND in cartilage degeneration. (A) The protein expression level of Collagen II, MMP13, EIF1, and JUND was examined by Wb. (B) Relative protein expression was quantified by densitometry. α-Tubulin was used as the internal control. (C) The protein expression level of EIF1, and JUND was examined by Wb after chondrocytes were transfected with distinct siRNAs for 72 h. α-Tubulin was used as the internal control. (D, E) Relative protein expression of EIF1 (D), and JUND (E) was quantified by densitometry. (F) The impact of EIF1 knockdown on the metabolism of ECM in chondrocytes. The protein expression level of Collagen II, and MMP13 was detected by Wb. Relative protein expression was quantified by densitometry. GAPDH was used as the internal control. (G) The impact of silencing JUND on the metabolism of ECM in chondrocytes. The protein expression level of Collagen II and MMP13 was detected by Wb. Relative protein expression was quantified by densitometry. GAPDH was used as the internal control. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Discussion

OA is a multifactorial degenerative disease accompanied by multiple complex mechanisms involved in its progression^[33]. Although significant advancements have been made in understanding the pathogenesis of OA, there is currently a lack of effective therapies to cure or relieve OA progression. Therefore, the development of novel biomarkers and therapeutic targets based on the molecular mechanisms is essential for the overall treatment of OA. Recently, m7G modification has been reported to function as a significant form of epigenetic regulation and to play critical roles in the development of numerous diseases. In the current study, we comprehensively investigated the biomarkers, predictive models, phenotypes, and immune infiltrates associated with m7G modification in OA. These findings suggest a novel strategy for OA diagnosis and treatment.

In this study, the peripheral blood of patients with OA was analyzed using various bioinformatic approaches, leading to the identification of four key m7G regulators: EIF1, JUND, NUDT16L1, and NCBP1. Remarkably, all of these biomarkers have the potential to be used as diagnostic biomarkers. Several key m7G regulators are involved in various BPs and diseases. For instance, it has been reported that EIF1 is involved in start codon recognition during eukaryotic translation initiation^[34,35]. The expression of EIF1 was found to be decreased in pancreatic ductal adenocarcinoma (PDAC), suggesting its critical role in translation during PDAC development^[36]. JUND, a member of the Jun family, is upregulated in cervical cancer, and knockdown of JUND inhibits the proliferation of HeLa cells by blocking the cell cycle in the G1-phase^[37]. It is shown that the expression of JUND was also increased in esophageal squamous cell carcinoma (ESCC) cells. Furthermore, JUND enhances the transcription of MAPRE2, thereby facilitating the proliferation and angiogenesis of ESCC cells^[38]. The NCBP1 protein is a component of the nuclear cap-binding protein complex, which binds to the 5′-end cap of pre-mRNAs to participate in RNA processing, nuclear export, and translation. It was found that NCBP1 was remarkably increased in lung cancer tissues and several lung cancer cell lines, which could promote cancer cell growth, wound healing ability, migration, and epithelial-mesenchymal transition through mechanical binding to the NCBP3 in nucleoplasm^[39]. Moreover, NCBP1 exhibited high expression levels in patients with diffuse large B-cell lymphoma (DLBCL) and demonstrated a positive correlation with an unfavorable prognosis. The upregulation of NCBP1 can promote the proliferation of DLBCL cells by regulating the c-MYC expression of c-MYC in a METTL3-dependent manner^[40]. To investigate the potential roles of these key regulators in OA, we conducted enrichment analyses to evaluate their involvement in various BPs and pathways. GO analysis suggested that these key m7G regulators were mainly enriched in RNA metabolism and processing, such as RNA catabolic processes, mRNA metabolic processes, and the regulation of translational initiation. KEGG analysis indicated that these key m7G regulators were mainly involved in RNA transport, IL-17 signaling pathway, and osteoclast differentiation, which have been implicated in cartilage degeneration^[41–43].

OA pathogenesis involves a range of pathological changes, including cartilage degeneration, synovitis, subchondral osteosclerosis, and osteophyte formation^[33]. The multitude of pathological alterations and the intricacy of pathogenesis poses challenges in the treatment of OA. The development of biomarkers and predictive models is believed to be beneficial for the early diagnosis and evaluation of therapeutic effects, thereby facilitating comprehensive treatment of patients with OA. To date, numerous biomarkers and predictive models have been identified from different perspectives, such as cuproptosis-related genes in the synovium^[44], DNA methylation-related genes in peripheral blood^[45], and RNA modification-related regulators in the subchondral bone^[11]. As an important pattern of RNA modification, the m7G modification has received increasing attention. Based on the expression of 4 m7G-related hub genes, SNUPN, METTL1, EIF4E2 and CYFIP1, Chen et al constructed a nomogram predictive model, and calibration curve analysis indicated that this predictive model functioned as an ideal predictive model for OA^[26]. Another predictive nomogram model was built based on the expression of EIF4E2, DCP2, SNUPN, and LARP1, which exhibited favorable efficiency in distinguishing early- and end-stage OA^[27]. Remarkably, both predictive models were mainly based on hub genes identified in the synovial tissues. In the present study, the dataset obtained from peripheral blood samples of 106 patients with OA and 33 healthy individuals was used to develop a more suitable model for clinical application. Based on the expression levels of the four key m7G regulators, we developed a predictive model that could function satisfactorily in forecasting the probability of OA. Moreover, a risk model was built using the LASSO analysis. Interestingly, OA patients had a higher risk score than healthy individuals, and patients in the high-risk group had an elevated risk score compared to patients in the low-risk group. Considering these findings, the predictive model and the risk model constructed using the expression data from the peripheral blood of key m7G regulators may be useful for the early and timely diagnosis of OA.

Nowadays, the paradigm of OA is evolving from a simplex mechanical disorder caused by cartilage wear to a systemic disease accompanied by changes in biomechanics, inflammation, and immune response^[8,46]. Immune cell infiltration by T cells, B cell subpopulations, and activated macrophages in synovial tissues is associated with knee pain and OA severity, indicating that both acquired and innate immunity may play a substantial role in OA pathogenesis^[16]. Sae-Jung et al confirmed that unhealthy aging could induce healthy aging of T-cells into a state of imbalance, leading to the occurrence and development of OA^[47]. A Mendelian Randomization study showed that CD4(+)CD25(+)T cells exert a protective effect against the progression of hip OA^[48]. Störch et al found that activated B cells obtained from healthy donors could lead to the conversion of non-arthritic synovial fibroblasts into a proinflammatory phenotype after co-culture with synovial fibroblasts of non-synovitic origin from OA patients^[49]. Furthermore, upregulation of TIMP-1, VEGF, and MMP-13 was found to be accompanied by CD8(+) T cells activation, thereby inducing cartilage degeneration^[50]. Recently, several studies have investigated the association between m7G modifications and immune cell infiltration. For instance, Lu et al demonstrated that the expression of multiple m7G genes was higher in COVID-19 patients than in non-COVID-19 patients, and that m7G-related cluster B showed higher immune infiltration^[51]. Methyltransferase 1 (METTL1), an important regulator of m7G modification, has been shown to play critical roles in modulating the immunosuppressive microenvironment of hepatocellular carcinoma^[52]. Moreover, METTL1 was highly expressed in primary and advanced prostate tumors, and knockdown of METTL1 promoted intratumoral infiltration of pro-inflammatory immune cells and the response to immunotherapy in prostate cancer preclinical models^[53]. It has been shown that two m7G regulators, NCBP2 and EIF4E3 could affect the immune context and response to immunotherapy by regulating their downstream gene expression in patients with head and neck squamous cell carcinoma^[54]. Liu et al found that multiple m7G modification-related genes are involved in the regulation of the tumor microenvironment, immune cell infiltration, response to immunotherapy, and patient prognosis in lung adenocarcinoma^[55]. In the current study, the ssGSEA results indicated that distinct risk groups, m7G phenotypes, and gene clusters showed different infiltrating features of the immune cells. Further analyses revealed significant associations between these key m7G regulators and the presence of multiple immunocytes. Taken together, m7G modification is closely associated with the immune microenvironment and m7G modification-related risk models, phenotypes, and gene clusters may help estimate the level of immune infiltration in patients with OA. However, further studies are needed to determine the role of m7G modification in OA pathogenesis from the perspective of immune homeostasis.

Cartilage degeneration is considered the core pathological change in OA and is one of the main causes of pain and disability among middle-aged and elderly individuals^[56]. Gaining deeper insight into the correlation between m7G modification and cartilage degeneration may contribute to a more comprehensive understanding of the mechanisms underlying OA. Hence, we further investigated the expression of key m7G regulators in cartilage using a single-cell analysis. Interestingly, we observed widespread EIF1 and JUND expression in all subgroups of chondrocytes, whereas NUDT16L1 and NCBP1 exhibited limited expression. Moreover, findings from an external validation dataset indicated downregulation of both EIF1 and JUND in OA cartilages. However, the expressions of NUDT16L1 and NCBP1 were not significantly different between the groups. Furthermore, ROC analysis confirmed that the AUC values of EIF1 and JUND were both greater than 0.7, indicating a significant discriminatory ability between OA and non-OA cartilages. Increasing evidence has confirmed a correlation between these two key m7G regulators and musculoskeletal disorders. A single-cell analysis conducted by Cherif et al revealed that EIF1 could potentially represent a distinct cell type within human intervertebral discs^[57]. Fu et al found that EIF1 expression was higher in fatty acid synthase-silenced osteosarcoma 143 B cells than in their parental cells^[58]. Nevertheless, EIF1 is downregulated in the human chondrosarcoma cell line HCS-2/8 when cultured under low-oxygen conditions^[59]. JUND is an activator protein-1 component and its deficiency enhances new bone formation, thereby reversing the low bone mass induced by estrogen depletion^[60]. It has been demonstrated that JUND could activate the transcription of lncRNA LOXL1-AS1, thereby regulating the downstream miR-423-5p/KDM5C axis and ultimately increasing the proliferation and inflammatory level of OA chondrocytes^[61]. To explore the potential roles of EIF1 and JUND in OA development, we collected normal human cartilage and conducted in vitro experiments to explore their association with cartilage degeneration. qRT-PCR results revealed a significant reduction in EIF1 and JUND expression in the OA cartilage. Moreover, the protein levels of both EIF1 and JUND decreased in TBHP-induced chondrocytes. Significantly, silencing EIF1 and JUND in chondrocytes led to upregulation of MMP13 and downregulation of collagen II. Moreover, suppression of EIF1 and JUND accelerated the cartilage degeneration in vivo. These findings indicate that both EIF1 and JUND may function as key therapeutic targets in the prevention of cartilage degeneration, which may provide some theoretical foundation for future clinical drug research in the treatment of OA.

However, the current study had some limitations. First, the findings relied primarily on an analysis of data obtained from public databases. However, it is important to note that the sample size of scRNA-seq and RNA-seq data derived from cartilage was relatively limited, which could potentially impact the validation of key m7G regulators in the cartilage. Second, the presence of confounding factors, such as age, sex, complications, and OA severity, in the GSE48556 dataset is inevitable and may introduce bias to the results. Moreover, some important m7G regulators may not have been chosen for analysis because of the heterogeneity of samples and differences in sequencing techniques among the datasets. Third, although multiple extensive and solid analyses were conducted, it is necessary to conduct further experiments to strengthen the conclusions of future studies. Fourth, future longitudinal cohorts for collecting blood samples from patients with different grades of OA for verification will greatly improve the integrity of this study. Despite these limitations, we believe that our study can serve as a valuable reference for further investigation of the role of m7G modifications in the pathogenesis of OA.

Conclusion

In summary, this study systematically investigated the potential implications of m7G modifications in the pathogenesis and progression of OA. Moreover, the screened key m7G regulators, including EIF1, JUND, NUT16L1, and NCBP1, exhibited favorable diagnostic values for OA and were closely associated with immune infiltration. In addition, both EIF1 and JUND may serve as novel therapeutic targets for cartilage degeneration. Collectively, these findings provide novel insights into the involvement of m7G modification in the molecular mechanisms of OA.

Ethical approval

This study was approved by the ethics committee of the Affiliated Hospital of Guizhou Medical University (No. 276 in 2022).

Consent

Not applicable.

Sources of funding

The study was funded by the Science and Technology Fund of Guizhou Science and Technology Department (QKH-ZK [2021] 391; QKH-ZK [2023] 344), the Science and Technology Fund of Guizhou Provincial Health Commission (gzwjkj2020-1-120; gzwkj2021-261), the Youth Fund cultivation program of National Natural Science Foundation of Affiliated Hospital of Guizhou Medical University (gyfynsfc-2021-12), and the Graduate Scientific Research Fund project of Guizhou (YJSKYJJ [2021] 157).

Author contributions

This study was conceived by H.Y. and X.N; H.S. and K.C. conducted bioinformatics analyses and experiments; Z.X. acquired the data from an online database; Y.Z. collected the cartilage samples; M.L. prepared the figures and chart drawings; X.N. supervised the implementation of this study; the funding was provided by H.S. and H.Y.; H.S. and K.C. prepared the original manuscript draft; and H.Y. and X.N. reviewed and edited the manuscript. The final version of the manuscript was approved by all authors.

Conflicts of interest disclosure

All authors declare that there is no competing interest.

Research registration unique identifying number (UIN)

Not applicable.

Guarantor

Xu Ning, and Hua Yang.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Data availability statement

The original data presented in this study are included in the paper. Moreover, the source data and methodology are listed in the Supplemental files.